---
title: "Netflix Titles"
author: "Antony Rono"
output:
flexdashboard::flex_dashboard:
orientation: rows
vertical_layout: fill
source_code: embed
editor_options:
chunk_output_type: console
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = FALSE, warning = FALSE, message = FALSE)
library(flexdashboard)
library(tidyverse)
library(tidytuesdayR)
library(plotly)
library(janitor)
library(magrittr)
library(lubridate)
library(scales)
library(ggsci)
theme_set(theme_light())
```
```{r Load data, include=FALSE}
# tt <- tt_load("2021-04-20")
#
# data <- tt$netflix_titles
data <- readr::read_csv('https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2021/2021-04-20/netflix_titles.csv')
netflix <- data %>%
clean_names() %>%
separate(duration, c("duration", "duration_units"), sep = " ", convert = TRUE) %>%
mutate(decade = 10*(release_year %/% 10))
skimr::skim(netflix)
```
```{r function to summarize movies, include=FALSE}
summarize_titles <- function(data){
data %>%
summarize(n = n(),
median_duration = median(duration),
median_year = median(release_year),
avg_duration = mean(duration),
) %>%
arrange(desc(n))
}
```
```{r function to summarize by count and percent, include=FALSE}
summarize_by_count_and_percent <- function(x, y){
netflix %>%
separate_rows({{x}}, sep = ",") %>%
separate_rows({{y}}, sep = ",") %>%
mutate(
{{x}} := str_trim({{x}}),
{{y}} := str_trim({{y}})
) %>%
filter(!is.na({{x}}), !is.na({{y}})) %>%
mutate({{x}} := fct_lump_n({{x}},10),
{{y}} := fct_lump_n({{y}},15)
) %>%
count({{x}}, {{y}}, name = "count", sort = TRUE) %>%
group_by({{x}}) %>%
mutate(pct = count/sum(count)) %>%
ungroup() %>%
mutate({{x}} := fct_reorder({{x}}, count, sum, .desc = FALSE),
{{y}} := fct_reorder({{y}}, count, sum, .desc = FALSE)
)
}
```
Overall {data-icon="fa-chart-line"}
=======================================================================
Row {data-height=250}
-----------------------------------------------------------------------
### Total Releases
```{r count of released}
valueBox(value = comma(nrow(data)), caption = "Total Releases", icon = "fa-user-plus", color = "green")
```
### Number of Countries That Have Been Produced In
```{r number of countries}
countries <- netflix %>%
separate_rows(country, sep = ",") %>%
mutate(country = str_trim(country)) %>%
distinct(country) %>%
nrow %>%
comma
gauge(countries, min = 0, max = 195, gaugeSectors(success = c(131, 195), warning = c(66,130), danger = c(0, 65)))
```
### Percentage of Movies and TV Shows
```{r proportion of movies and tv shows}
netflix %>%
count(type) %>%
plot_ly(labels = ~ type, values = ~ n) %>%
add_pie(hole = 0.6)
```
Row {data-height=750}
-----------------------------------------------------------------------
### Overall Trend in Productions in the Last Two Decades
```{r production by year - stacked bar chart}
type_year_count_df <- netflix %>%
count(type, release_year) %>%
filter(release_year > 2000)
p <- type_year_count_df %>%
ggplot(aes(release_year, n)) +
geom_col(aes(fill = type)) +
geom_line(data = type_year_count_df %>% filter(type == "Movie"), color = "red1", size = 1) +
geom_line(data = type_year_count_df %>% filter(type == "TV Show"), color = "cyan4", size = 1) +
scale_y_continuous(expand = c(0,0)) +
scale_x_continuous(expand = c(0,0), n.breaks = 11) +
labs(y = "Count")
ggplotly(p)
```
### Relative Trend in TV Shows and Movies
```{r}
p <- netflix %>%
count(type, release_year) %>%
group_by(release_year) %>%
pivot_wider(names_from = type, values_from = n, values_fill = 0) %>%
clean_names() %>%
mutate(pct_tv_shows = tv_show/sum(movie, tv_show),
pct_movies = 1 - pct_tv_shows) %>%
pivot_longer(cols = starts_with("pct"), names_to = "type", values_to = "pct") %>%
mutate(type = ifelse(str_detect(type, "movies"), "Movie", "TV Show")) %>%
ggplot(aes(release_year, pct, color = type))+
geom_point() +
geom_line(arrow = arrow(angle = 15, ends = "last", type = "closed")) +
scale_y_continuous(labels = percent, expand = c(0,0), limits = c(0,1.05)) +
scale_x_continuous(expand = c(0,0), n.breaks = 15, limits = c(1925, 2022)) +
labs(y = "Proportion of all releases")
ggplotly(p)
```
Country {data-orientation=columns data-icon="fa-location-arrow"}
=======================================================================
Column
-----------------------------------------------------------------------
### Total Number of Productions by Country
```{r total production by countr - worldmap}
library(highcharter)
library(maps)
prod_country <- netflix %>%
separate_rows(country, sep = ",") %>%
mutate(country = str_trim(country)) %>%
filter(str_detect(country, "")) %>%
count(country) %>%
mutate(country = case_when(
str_detect(country, "Germany") ~ "Germany",
country == "United Kingdom" ~ "UK",
country == "Soviet Union" ~ "Russia",
country == "Vatican City" ~ "Vatican",
TRUE ~ country
)) %>%
mutate(`iso-a3`= coalesce(iso3166$a3[match(country,iso3166$sovereignty)], iso3166$a3[match(country , iso3166$ISOname)]))
hcmap(
map = "custom/world-highres3", # high resolution world map
data = prod_country, # name of dataset
joinBy = "iso-a3",
value = "n",
showInLegend = FALSE, # hide legend
nullColor = "#DADADA",
download_map_data = TRUE
) %>%
hc_mapNavigation(enabled = FALSE) %>%
hc_legend("none") #%>%
#hc_title(text = "Production By Country") # title
```
Column
-----------------------------------------------------------------------
### Breakdown by Country and Type
```{r production by country and type-worldmap }
p <- netflix %>%
separate_rows(country, sep = ",") %>%
mutate(country = str_trim(country)) %>%
filter(!is.na(country)) %>%
count(type, country = fct_lump(country, 15), sort = TRUE, name = "count") %>%
mutate(country = fct_reorder(country, count, sum),
) %>%
ggplot(aes(country, count, fill = type)) +
geom_col() +
scale_y_continuous(expand = c(0,0)) +
coord_flip() +
labs(y = "Count")
ggplotly(p)
```
### Breakdown By Ratings
```{r production by country and ratings}
p <- summarize_by_count_and_percent(country, rating) %>%
ggplot(aes(country, pct, fill = rating))+
geom_col() +
scale_y_continuous(expand = c(0,0), labels = percent) +
coord_flip() +
labs(y = "proportion of all releases")
ggplotly(p)
```
Genre {data-orientation=columns data-icon="fa-code"}
=======================================================================
Column
-----------------------------------------------------------------------
### Breakdown by Type
```{r production by genre}
p <- netflix %>%
separate_rows(listed_in, sep = ",") %>%
mutate(listed_in = str_trim(listed_in)) %>%
count(type, listed_in) %>%
mutate(listed_in = fct_reorder(listed_in, n, sum)) %>%
filter(!is.na(listed_in)) %>%
slice_max(n, n =30) %>%
ggplot(aes(listed_in, n, fill = type)) +
geom_col() +
scale_y_continuous(expand = c(0,0)) +
coord_flip() +
labs(y = "Count",
x = "Genre") +
facet_wrap(~type, scales = "free", ncol = 2) +
theme(legend.position = "none")
ggplotly(p)
```
Column
-----------------------------------------------------------------------
### Breakdown By Rating
```{r production by genre and ratings}
p <- summarize_by_count_and_percent(listed_in, rating) %>%
ggplot(aes(listed_in, pct, fill = rating))+
geom_col() +
scale_y_continuous(expand = c(0,0), labels = percent) +
coord_flip() +
labs(y = "proportion of all ratings",
x = "genre")
ggplotly(p)
```
Duration {data-icon="fa-calendar-check"}
=======================================================================
Row
-----------------------------------------------------------------------
### Trend of Length of Movies Overtime
```{r duration of movies over the years}
p <- ggplot() +
geom_boxplot(data = netflix %>% filter(type == "Movie") , aes(decade, duration,group = decade), outlier.colour = "red", outlier.shape = 1) +
geom_line(data = netflix %>% filter(type == "Movie") %>% group_by(decade) %>% summarise(avg_length = median(duration)) %>% ungroup(), aes(decade, avg_length),size = 1, color = "darkorchid4")+
scale_x_continuous(n.breaks = 15)
ggplotly(p)
```
Row
-----------------------------------------------------------------------
### Breakdown By Country
```{r duration of movies by country}
p <- netflix %>%
separate_rows(country, sep = ",") %>%
mutate(country = str_trim(country)) %>%
group_by(type, country = fct_lump(country, 20)) %>%
summarize_titles() %>%
filter(type == "Movie") %>%
mutate(country = fct_reorder(country, median_duration)) %>%
filter(!is.na(country)) %>%
ggplot(aes(country, median_duration, fill = country)) +
geom_col() +
scale_y_continuous(expand = c(0,0)) +
theme(legend.position = "none") +
coord_flip() +
labs(y = "Duration (minutes)",
x = "Genre")
ggplotly(p)
```
### Breakdown by Genre
```{r duration of movies by genre}
p <- netflix %>%
separate_rows(listed_in, sep = ",") %>%
mutate(listed_in = str_trim(listed_in)) %>%
group_by(type, listed_in) %>%
summarize_titles() %>%
filter(type == "Movie") %>%
filter(listed_in != "Movies") %>%
mutate(listed_in = fct_reorder(listed_in, median_duration)) %>%
filter(!is.na(listed_in)) %>%
ggplot(aes(listed_in, median_duration, fill = listed_in)) +
geom_col() +
scale_y_continuous(expand = c(0,0)) +
theme(legend.position = "none") +
coord_flip() +
labs(y = "Duration (minutes)",
x = "Genre")
ggplotly(p)
```
Ratings {data-icon="fa-thumbs-up"}
=======================================================================
Row
-----------------------------------------------------------------------
### Trend of Ratings Overtime
```{r ratings overtime trend}
p <- netflix %>%
filter(!is.na(rating)) %>%
mutate(rating = fct_lump_n(rating,5)) %>%
count(rating, decade, name = "count", sort = TRUE) %>%
mutate(rating = fct_reorder(rating, count, sum, .desc = TRUE)) %>%
ggplot(aes(decade, count, color = rating))+
geom_point()+
geom_line() +
scale_x_continuous(n.breaks = 15)
ggplotly(p)
```
### Ratings By Country
```{r ratings by country}
p <- summarize_by_count_and_percent(rating, country) %>%
ggplot(aes(rating, pct, fill = country))+
geom_col() +
coord_flip() +
scale_y_continuous(labels = percent)+
#scale_fill_discrete(palette = "Spectral")+
labs(y = "proportion of all releases")
ggplotly(p)
```
Row
-----------------------------------------------------------------------
### Countries that have mature content (R/TV-MA)
```{r countris by mature content}
p <- netflix %>%
separate_rows(country, sep = ",") %>%
mutate(country=str_trim(country)) %>%
filter(!is.na(rating), !is.na(country)) %>%
group_by(type, country = fct_lump(country, 10)) %>%
summarize(n_mature = sum(rating %in% c("R", "TV-MA", "NC-17")),
n = n(),
.groups = "drop"
) %>%
mutate(pct_mature = n_mature / n,
conf_low = qbeta(0.025, n_mature + 0.5, n - n_mature + 0.5),
conf_high = qbeta(0.975, n_mature + 0.5, n - n_mature + 0.5),
) %>%
ggplot(aes(pct_mature, country, color = type))+
geom_point(aes(size = n)) +
geom_errorbar(aes(xmin = conf_low, xmax =conf_high ), width = 0.5) +
scale_x_continuous(labels = percent, n.breaks = 4) +
labs(x = "% of releases that are Mature (R/TV-MA/NC-17 rated)",
size = "No. of releases")
ggplotly(p)
```
### Rating By Genre
```{r ratings by genre}
p <- summarize_by_count_and_percent(rating, listed_in) %>%
ggplot(aes(rating, pct, fill = listed_in))+
geom_col() +
coord_flip() +
scale_y_continuous(labels = percent)+
#scale_fill_discrete(palette = "Spectral")+
labs(y = "proportion of all releases",
fill = "Genre")
ggplotly(p)
```
Word Frequency {data-orientation=columns data-icon="fa-globe"}
=======================================================================
Column {.tabset}
-----------------------------------------------------------------------
### Frequency of words in title -Wordcloud
```{r frequency of words in title - wordcloud}
library(tidytext)
library(wordcloud2)
library(webshot)
library(htmlwidgets)
word_title_freq <- netflix %>%
unnest_tokens(word, title, drop = FALSE) %>%
anti_join(stop_words, by = "word") %>%
filter(str_length(word) >2) %>%
count(word) %>%
arrange(desc(n)) %>%
rename(freq = n) %>%
filter(freq > 1)
#slice_max(n, n = 70)
graph <- wordcloud2(word_title_freq, size=0.8, terrain.colors(length(word_title_freq$word)), shape = "circle", backgroundColor = "black")
#webshot::install_phantomjs()
# save it in html
saveWidget(graph,"netflix_titles_wordcloud.html",selfcontained = F)
# and in png or pdf
webshot("netflix_titles_wordcloud.html","netflix_titles_wordcloud.png", delay =8, vwidth = 480, vheight=480)
graph
```
### Most Common words in titles for each type
```{r most common words in description}
library(snakecase)
library(tidylo)
words_unnested <- netflix %>%
unnest_tokens(word, description, drop = FALSE) %>%
anti_join(stop_words, by = "word") %>%
filter(str_length(word) >2)
p <- words_unnested %>%
count(type, word) %>%
bind_log_odds(type, word, n) %>%
arrange(desc(log_odds_weighted)) %>%
group_by(type) %>%
slice_max(log_odds_weighted, n=15) %>%
ungroup() %>%
mutate(word = fct_reorder(word, log_odds_weighted)) %>%
ggplot(aes(log_odds_weighted, word, fill = type)) +
geom_col() +
scale_x_continuous(expand = c(0,0))+
theme(legend.position = "none")+
facet_wrap(~type, scales = "free_y") +
labs(x = "Log odds ratio, weighted by uninformative Dirichlet prior")
ggplotly(p)
```
Column
-----------------------------------------------------------------------
### Cluster of words that tend to appear together in movie/tv description
```{r clustering of words}
library(widyr)
library(tidygraph)
library(ggraph)
words_unnested <- netflix %>%
unnest_tokens(word, description, drop = FALSE) %>%
anti_join(stop_words, by = "word") %>%
filter(str_length(word) >2)
words_unnested %>%
distinct(type, title, word ) %>%
add_count(word, name = "word_total") %>%
arrange(desc(word_total)) %>%
filter(word_total >=40) %>%
pairwise_cor(word, title, sort = TRUE) %>%
filter(correlation > 0.1) %>%
igraph::graph_from_data_frame() %>%
#ggraph(layout = "igraph", algorithm = 'kk') +
ggraph(layout = "fr")+
geom_edge_link(aes(alpha = correlation)) +
geom_node_point(col = "red") +
#geom_node_text(aes(label = name), hjust = 1, vjust =1 , check_overlap = TRUE)
geom_node_text(aes(label = name), repel = TRUE) +
theme(legend.position = "none")+
scale_color_brewer(palette = "Set1") +
scale_fill_brewer(palette = "Set1")
```
